New - Live monitoring now understands containers: inside Docker, Podman and Kubernetes containers, CPU, memory, disk and uptime show the container's own figures. - Expand live monitoring to see the processes using the most resources. - Network link tests show progress: each result appears as soon as it's ready, with an estimate of the time remaining. - Custom AI providers can set reasoning effort and extra request headers. Improvements - Redesigned connection list, connection details, logs and network monitoring pages. - If a connection drops, the monitoring card keeps its last readings and marks the connection as interrupted. - A newly saved connection is now selected and scrolled into view. - Click an empty part of the path bar to type a path directly. - Default AI models are updated to the latest versions, and failed AI requests now show the provider's reason. - The Hide Address setting now also applies to network monitoring. Bug Fixes - Fixed symbols, punctuation and repeated keys occasionally being dropped in the terminal when using Chinese or Japanese input methods. - When a connection through a jump host fails, the error now says whether the jump host or the target server is the problem. - Fixed servers without Docker showing a placeholder container in the container list. - Fixed errors when using newer Claude and Gemini models.